gtmskills.com

Reference file

Research protocol

research-protocol.md

Per-competitor research protocol

Run this protocol once per competitor, in parallel threads when researching several. Each thread gets: the competitor's name, its domain, and its previous positioning snapshot (or "first run — no comparison available"). Budget roughly a dozen research calls per competitor — favor extraction quality over quantity.

Phase 0 — Site discovery (always first)

Map the site's actual structure (sitemap fetch or shallow crawl, up to ~200 URLs) before extracting anything. Guessed paths 404; discovered paths don't.

From the URL list, identify:

  • Homepage — always extract.
  • Features/product page — paths containing /features, /product, /platform, /solutions, /why-us, /capabilities. Pick the single most relevant.
  • Pricing page — /pricing, /plans, /get-started.
  • Blog index — /blog, /resources, /content, /insights.
  • Case studies — /case-study, /customers, /stories.

Record which pages exist and which don't. Absences are positioning signals in their own right: no public pricing page suggests enterprise/sales-led motion; no blog suggests the company isn't competing on content. Note notable subdomains too (docs., api., community.) — a docs subdomain signals a developer audience regardless of what the homepage claims.

If the site blocks sitemap/crawl access, fall back to extracting the homepage directly and trying the common paths above by hand.

Phase 1 — Page extraction (run simultaneously)

Extract each discovered page as clean text/markdown:

  1. Homepage → hero copy, tagline, primary CTA, value propositions, top-level nav structure.
  2. Features page → feature categories, branded feature names, emphasis, explicit differentiation claims ("unlike X", "the only platform that…"). Skip if none found.
  3. Pricing page → tier names, pricing model, entry price, feature gating, target audience per tier. Skip if none found.

If an extraction returns garbage, retry once; if still empty, record "page not accessible" and continue.

Phase 2 — Blog and content (run simultaneously)

Three searches:

  1. Search within the competitor's domain for its blog index and recent posts (~10 results).
  2. Search the live web for the competitor's name plus blog/content/resources terms, windowed to the period since the last run (~10 results). Verify actual publish dates from the extracted content — search-result dates lie.
  3. Search within the competitor's domain for case studies, customer stories, and testimonials (~5 results).

Some companies don't blog. Record "no active blog detected" and lean on page-based positioning instead — don't invent content themes from nothing.

Phase 3 — Deep extraction (if posts found)

Extract the 2–3 most recent blog posts in full for theme analysis: what story does this company tell repeatedly, who is it aimed at, do they name competitors or position against a category, what keywords do titles and headings chase.

Output — return exactly these sections

SITE STRUCTURE

  • Pages found / pages missing (with the implication of each absence)
  • Subdomains
  • Structure signals — what the architecture reveals about positioning

HOMEPAGE POSITIONING

  • Tagline (exact text) · Hero message · Primary CTA (button text + implied action)
  • Value props, each verbatim
  • Navigation structure (top-level items — what they choose to emphasize)
  • Target audience signals (who the copy speaks to)

FEATURES PAGE

  • Page URL · Key feature categories · Differentiation claims · Branded feature names
  • Missing/notable — conspicuously absent or prominently highlighted capabilities
  • (No features page → write "No public features page — [implication]")

PRICING PAGE

  • Page URL · Model (per-seat / usage / flat / freemium / custom) · Tier names
  • Entry price if visible · Enterprise signal ("contact sales", custom tier) · Feature gating
  • (No pricing page → write "No public pricing — likely enterprise/sales-led")

BLOG & CONTENT

  • Last 5–10 posts with titles and dates · Primary recurring themes
  • Content types (blog, case study, whitepaper, webinar) · Publishing cadence estimate
  • Audience focus (developer / executive / practitioner)

SOCIAL PROOF

  • Customer logos/names if public · Testimonial themes (what customers praise)
  • Case study focus (industries, use cases, outcomes highlighted)

CHANGES FROM PREVIOUS SNAPSHOT

  • Specific diffs only: "Tagline changed from '[old]' to '[new]'", "New feature category added: [name]", "Pricing model shifted from [old] to [new]", "New blog theme emerging: [topic]", "Page added/removed: [url] — [significance]"
  • First run → "First run — no comparison available"